Bernd,
it would be nice if you could share with us what you had to change to
make things work with Float128. Ideally, you would fork the
corresponding repository and create one or several merge requests like
described here:
I was about to make the merge request, but apparenly I can not fork the
dumux repo until I am given some privileges in the project. I pushed the
"request access" button and will see how it goes.
With "seems to work", do you mean that the irregularities vanish?
No, I just meant that the code compiles and runs. I am yet to see
whether I am able to set the residual criterion that makes the
irregularities sufficiently small for the time of simulation, and, at
the same time, allows the method to converge at all. That turned out to
be impossible with regular 64-bit floats. I will continue attempts.
If you want to quantify the influence of physical parameters on the
fingers, you have to apply measures that don't depend on the numerical
setting like grid size or solver thresholds. Which is probably
something like onset time, finger length and things like that, but I'm
certainly no expert here.
Oh, finger length certainly depends on grid size in my simulations. With
the grids I was able to try, anyway. The finer the grid, the thinner and
faster are the fingers. Maybe I have not yet reached the grid resolution
at which it is no longer the case. In fact, finger length seems to be
depending on everything (except maybe the time step, though I am not
100% sure).
Best regards,
Dmitry
_______________________________________________
DuMux mailing list
[email protected]
https://listserv.uni-stuttgart.de/mailman/listinfo/dumux
